Ecological Engineering 120 (2018) 432–440

Contents lists available at ScienceDirect

Ecological Engineering
journal homepage: www.elsevier.com/locate/ecoleng

Assessing the effectiveness of riparian restoration projects using Landsat and T

precipitation data from the cloud-computing application ClimateEngine.org
⁎
Mark B. Hausnera, , Justin L. Huntingtona, Caroline Nashb, Charles Mortonc, Daniel J. McEvoyd,
David S. Pilliode, Katherine C. Hegewischf, Britta Dauderta, John T. Abatzoglouf, Gordon Grantb
a
Division of Hydrologic Sciences, Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512, USA
b
College of Earth, Ocean and Atmospheric Sciences, Oregon State University, 3200 SW Jefferson Way, Corvallis, OR 97331, USA
c
Division of Earth and Ecosystem Sciences, Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512, USA
d
Division of Atmospheric Sciences, Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512, USA
e
U.S. Geological Survey, Forest and Rangeland Ecosystem Science Center, 970 Lusk St., Boise, ID 83706, USA
f
Department of Geography, University of Idaho, 875 Perimeter Dr., Moscow, ID 83844, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Riparian vegetation along streams provides a suite of ecosystem services in rangelands and thus is the target of
In-stream structures restoration when degraded by over-grazing, erosion, incision, or other disturbances. Assessments of restoration
Landsat effectiveness depend on defensible monitoring data, which can be both expensive and difficult to collect. We
NDVI present a method and case study to evaluate the effectiveness of restoration of riparian vegetation using a web-
Riparian vegetation
based cloud-computing and visualization tool (ClimateEngine.org) to access and process remote sensing and
Stream restoration
climate data. Restoration efforts on an Eastern Oregon ranch were assessed by analyzing the riparian areas of
Climate Engine
Google Earth Engine four creeks that had in-stream restoration structures constructed between 2008 and 2011. Within each study
Precipitation area, we retrieved spatially and temporally aggregated values of summer (June, July, August) normalized dif-
Remote sensing ference vegetation index (NDVI) and total precipitation for each water year (October-September) from 1984 to
2017. We established a pre-restoration (1984–2007) linear regression between total water year precipitation and
summer NDVI for each study area, and then compared the post-restoration (2012–2017) data to this pre-re-
storation relationship. In each study area, the post-restoration NDVI-precipitation relationship was statistically
distinct from the pre-restoration relationship, suggesting a change in the fundamental relationship between
precipitation and NDVI resulting from stream restoration. We infer that the in-stream structures, which raised
the water table in the adjacent riparian areas, provided additional water to the streamside vegetation that was
not available before restoration and reduced the dependence of riparian vegetation on precipitation. This ap-
proach provides a cost-effective, quantitative method for assessing the effects of stream restoration projects on
riparian vegetation.

1. Introduction from stream incision or other forms of disconnection between streams

Riparian areas provide critical ecosystem services and are subject to
degradation from natural and anthropogenic processes, especially in
arid and semi-arid rangelands where water is frequently limiting. A
riparian area can cool stream water through shading by streamside
vegetation, filter surface and ground water, and otherwise act as an
interface between uplands and waters (Gregory et al., 1991; Naiman
and Decamps, 1997). Riparian areas also support much of a region’s
biodiversity (Naiman et al., 1993; Sabo et al., 2005). These ecosystem
services, however, are sensitive to herbivory, disturbance (e.g., flood,
fire), and human activity. Degradation of riparian areas also can result
and their floodplains (NRC, 2002).
Degraded riparian areas are common targets of restoration projects
(Goodwin et al., 1997; Bernhardt et al., 2005), but evaluating restora-
tion effectiveness has long been a weakness in the field (Kondolf, 1995;
Walker et al., 2007; Gonzalez et al., 2015). Riparian restoration projects
are frequently underfunded (Ruiz-Jaen and Aide, 2005; Bernhardt
et al., 2007) and few projects budget for pre- and post-project mon-
itoring (Bernhardt et al., 2005; Gonzalez et al., 2015). Long-term
monitoring efforts are even rarer (Alexander and Allan, 2007; Gonzalez
et al., 2015) and monitoring data that do exist are seldom used to assess
the effectiveness of individual projects (Alexander and Allan, 2007).

⁎
Corresponding author.
E-mail address: mark.hausner@dri.edu (M.B. Hausner).

https://doi.org/10.1016/j.ecoleng.2018.06.024
Received 6 February 2018; Received in revised form 25 May 2018; Accepted 27 June 2018
0925-8574/ © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
M.B. Hausner et al.
Fig. 1. General study area of Silvies Valley Ranch, and tributary creeks and study riparian areas.
The development of metrics and standards for evaluating restora-
tion effectiveness (i.e., assessing whether or not restoration objectives
were met) has proven difficult (Bernhardt et al., 2007; Kondolf et al.,
2007) and rigorous evaluations have been even more difficult to im-
plement (Alexander and Allan, 2007; Gonzalez et al., 2015). Both
spatial (Hobbs and Norton, 1996; Gonzalez del Tanago and García de
Jalón, 2006; Aguiar et al., 2011) and temporal (Kondolf and Micheli,
1995; Trowbridge, 2007) scales prove challenging in evaluating effects
of restoration work. Stream or riparian restoration projects are gen-
erally implemented on a local scale (e.g., a meander bend of a river)
(Lake et al., 2007). However, environmental or ecological effects of
restoration often can be seen outside of project boundaries, including
upstream and downstream reaches and streamside areas (Aguiar et al.,
2011). Riparian vegetation in particular is characterized by hydrophilic
plants and thus inextricably linked to regional and local scale climate
and hydrology (Stromberg et al., 2007; Orellana et al., 2012; Boudell
et al., 2015). Riparian vegetation is strongly influenced by both long-
itudinal and lateral surface and subsurface connectivity to the flood-
plain (Gonzalez del Tanago and García de Jalón, 2006; Lake et al.,
2007). These linkages are even more pronounced in arid environments
such as rangelands (Patten, 1998; Perry et al., 2012).
433
Ecological Engineering 120 (2018) 432–440
Many restoration projects are assessed through metric trajectories,
or changes over time (Palmer et al., 2005; Gonzalez et al., 2015), yet
these evaluations require both baseline pre-restoration data and long-
term monitoring of the completed project. Although Kondolf & Micheli
(1995) recommend both a historical study of pre-restoration conditions
and a minimum of 10 years of post-restoration monitoring, Gonzalez
et al. (2015) found that just 16% of projects monitoring trajectories
included pre-restoration data, and only 22 of 169 total projects re-
viewed included more than six years of post-restoration monitoring.
The long-term effects of restoration will seldom be seen in such short
time frames (Trowbridge, 2007), especially in projects intended to
foster natural ecological processes like succession (Walker et al., 2007),
plant community development (Weisberg et al., 2013), and resilience
from natural disturbance (e.g., beaver, flood).
Satellite remote sensing has been used extensively to assess wetland
and riparian vegetation conditions at regional and local scales (Ozesmi
and Bauer, 2002; Goetz, 2006; Smith et al., 2014; Lawley et al., 2016).
Long-term remote sensing of riparian vegetation requires satellite ob-
servations of sufficient history and spatial resolution to characterize
both baseline conditions and trajectories with respect to natural and
anthropogenic change agents, such as climate, hydrology, and land
M.B. Hausner et al.
Fig. 2. Delineated riparian study areas. (a) Upper (left) and Lower (right) reaches of Camp Creek. (b) Flat Creek (larger polygon) and Jack Andy Creek (smaller
polygon).
management (Dawson et al., 2016). Landsat is ideally suited to meet
these requirements due to its extensive and continuous archive (30+
years), overpass frequency (8–16 day), and spatial resolution (30 m
pixel size) (Gutman and Masek, 2012; Huntington et al., 2016).
Since opening the Landsat archive, ecological applications of
Landsat data have exponentially increased (Wulder et al., 2012;
Kennedy et al., 2014). Software platforms such as Google Earth Engine
(Gorelick et al., 2017) now provide remote sensing and climate data
archive access and massively parallel cloud-computing capabilities.
This cloud-based access and processing has led to significant advance-
ments and data discoveries related to high resolution land cover and
water extent mapping, as well as visualizations that would not be
otherwise possible (Hansen et al., 2013; Pekel et al., 2016; Kennedy
et al., 2018). These new capabilities and advancements have changed
the paradigm of remote sensing and ecosystem monitoring over long
time histories and at high resolution.
Here we present a method and case study to assess the effectiveness
of riparian restoration using a freely available, on-demand, cloud-
computing web application to access, process, and download 30+ years
of Landsat and gridded precipitation data for restoration locations. We
compared statistical relationships between riparian vegetation vigor
and precipitation data for both pre- and post-restoration periods. The
approach we present overcomes some of the barriers in accessing large
geospatial datasets and provides a simple but rigorous quantitative
approach that can be used by scientists, practitioners, and managers to
434
Ecological Engineering 120 (2018) 432–440
assess the effectiveness of a restoration project for improving riparian
vegetation.
2. Materials and methods
2.1. Study site
This study focuses on streamside areas adjacent to three streams
tributary to the Silvies River, located on the Silvies Valley Ranch, in
eastern Oregon (Fig. 1). Streamside areas were identified as areas that
had the potential for inundation given their proximity to the channel,
slope, and topography (e.g., valley form). These areas were likely to
support riparian vegetation following restoration. Each stream had a
number of artificial beaver dams (ABDs) installed between 2008 and
2011, but none had pre-restoration vegetation or hydrologic data col-
lected and no systematic post-restoration data were collected (Davee
et al., 2017). ABDs are densely spaced, low-head weirs constructed from
rock and other materials that pond water up to the valley floor, thereby
raising water tables, slowing the movement of water through the
landscape, increasing valley bottom storage and shallow groundwater
levels, and promoting surface and groundwater interactions within re-
storation areas (Pilliod et al., 2018). Such changes ultimately lead to
replacement of upland shrubland vegetation (i.e., typical of sagebrush
steppe) with riparian vegetation and grasses, as well as increases in
vegetation vigor and evapotranspiration (Loheide and Gorelick, 2005;
M.B. Hausner et al.
Essaid and Hill, 2014).
The Silvies Valley Ranch encompasses the majority of a large allu-
vial valley, as well as some steeper, adjacent uplands in the headwater
reaches of the Silvies River, which drains to Malheur Lake. The land-
owner purchased the property in 2007 with the intent to establish a
sustainable grass-fed cattle grazing operation and eco-tourism resort.
Eager to return sparsely vegetated valley floors to productive grass-
lands, as well as to offer fishing and bird-watching to future guests, the
landowners set out to address the pervasive channel incision that had
dried upland valley floors, converting what was thought to have been
herbaceous wet meadows into sparse, upland species (e.g. Artemisia
spp., Chrysothamnus spp., Purshia tridentata, Festuca idahoensis).
To reverse this conversion, the landowners built 376 ABDs in six
intermittent tributary drainages over the course of six years. The ear-
liest in-stream structures were installed in the winter of 2008, and in-
stallation in the four study areas continued through 2011. The struc-
tures are maintained periodically by both the landowner and the
resident beaver populations, with the majority still operating as initially
designed. In the valley bottoms on either side of the structures, sage-
brush and other upland species were mechanically removed, and have
been replaced by a mix of grasses. Cattle are now regularly grazed in
the valley bottoms alongside the treated areas, and the areas are hayed
seasonally. The landowners report large increases in hay production
since installing the structures (Davee et al., 2017). Whereas the pre-
installation channels had been incised up to 5 m and went dry most
years, the post-installation channel is perennially wet and the water
level in some channels has risen by more than 3.5 m.
2.2. Study areas and data retrieval
Riparian areas for this study were selected based on the length of
time since the installation of the ABDs, with the most recent ABDs in-
stalled at least five years previously. After a field visit to the ranch to
identify potential study areas, we selected four study areas: Upper
Camp Creek, Lower Camp Creek, Flat Creek, and Jack Andy Creek
(Fig. 2). In each study area, a riparian area polygon was delineated on
the basis of high resolution background imagery (e.g. National Agri-
culture Imagery Program).
Analyses were based on the relationship between the satellite-de-
rived normalized difference vegetation index (NDVI) and precipitation.
NDVI is a measure of pixel greenness, or vegetation photosynthetic
potential. NDVI has been used as a proxy for ecosystem performance in
forests (Wylie et al., 2008, 2014) and rangelands (Wylie et al., 2012;
Rigge et al., 2013b), and changes in NDVI over time have been used to
examine the effect of best management practices on riparian vegetation
in rangelands (Rigge et al., 2013a). In the Great Basin, NDVI is used
extensively for quantifying vegetation vigor, plant cover, and con-
sumptive water use of groundwater dependent vegetation (McGwire
et al., 2000; Devitt et al., 2011; Huntington et al., 2016; Carroll et al.,
2017). Summer (June, July, and August; JJA) NDVI was chosen to
maximize the vegetation signal derived from shallow groundwater
(Dawson and Pate, 1996; Huntington et al., 2016), whereas water year
precipitation was chosen because it is a good indicator of shallow
groundwater levels and groundwater discharge (i.e. baseflow) during
this period (Huntington and Niswonger, 2012; McEvoy et al., 2012;
Abatzoglou et al., 2014).
We used Climate Engine (ClimateEngine.org), a freely available
remote sensing and climate cloud-computing application (Huntington
et al., 2017) powered by Google Earth Engine (Gorelick et al., 2017), to
process and download spatially and temporally averaged Landsat 4, 5,
7, and 8 derived annual NDVI values for each polygon from 1984 to
2017. NDVI was computed within Climate Engine using U.S. Geological
Survey (USGS) Landsat at-surface reflectance product collections hosted
by Google Earth Engine (USGS, 2018a,b). Climate Engine automatically
applies cloud masks provided by Landsat at-surface reflectance collec-
tions (Zhu and Woodcock, 2012; Foga et al., 2017) for data masking.
435
Ecological Engineering 120 (2018) 432–440
Spatially and temporally averaged NDVI values were computed within
Climate Engine and downloaded for respective polygons as median
values for Landsat images acquired in the summer (JJA) of each year.
Daily precipitation data from METDATA (Abatzoglou, 2013) was
aggregated to water year totals (October 1– September 30) and
downloaded for each study area via Climate Engine. METDATA is a
hybrid of daily North American Data Assimilation System (NLDAS)
(Mitchell et al., 2004; Xia et al., 2012) and monthly Parameter Re-
gression on Independent Slopes Model (PRISM) data (Daly et al., 2008),
and is available at a 4 km grid resolution. Two of the study areas were
contained entirely within a single METDATA grid cell; in those cases,
the total precipitation over the water year for respective cells were
used. The Upper and Lower Camp Creek study areas each spanned two
different METDATA cells, and the mean water year precipitation over
the two cells was used for these study areas.
2.3. Statistical analyses
Using the pre-restoration (1984–2007) NDVI and precipitation data,
we examined the relationship between precipitation and NDVI using
linear regression to determine a best-fit regression line for each study
area. Using MATLAB’s Curve Fitting Toolbox (MathWorks, Natick,
Massachusetts), we determined the 95 percent confidence intervals, the
95% prediction interval, the R2, F-statistic, and p-value for the linear
regression. We then compared post-restoration data (2012–2017) to
pre-restoration (1984–2007) regressions. If the post-restoration data fit
the pre-restoration relationship, the residuals between the post-re-
storation observations and the pre-restoration regression would be
normally distributed with a mean of zero. We tested this hypothesis
with a single-sample t-test on post-restoration residuals. This test shows
whether the post-restoration relationship between NDVI and pre-
cipitation is statistically significantly different from the pre-restoration
relationship.
To provide control areas, we performed analyses on Flat Creek (an
ephemeral and intermittent stream) and Lower Camp Creek (a per-
ennial stream). Of the four areas evaluated, Flat Creek is the most de-
pendent on precipitation for streamflow and vegetative growth,
whereas Camp Creek is the least dependent. In these control analyses,
the pre-restoration data were split into two datasets: an early pre-re-
storation (1984–2002) and a late pre-restoration (2003–2007). The
same analyses were run on these two pre-restoration datasets, estab-
lishing a statistical relationship between precipitation and NDVI for the
early data and using the residuals for the later pre-restoration data to
assess whether those later pre-restoration data fit the earlier pre-re-
storation relationship. Following the control, the post-restoration re-
siduals (2012–2017) were also tested against the early pre-restoration
data. All analyses were conducted in MATLAB using a 0.05 alpha level
as a basis for statistical significance.
2.4. Restoration assessment
We used the statistical analyses to assess whether the riparian re-
storation was successful in each study area. Fig. 3 provides a conceptual
illustration using synthetic data. If NDVI were entirely dependent on
precipitation, the pre-restoration linear regression would perfectly
predict the value of NDVI in response to a year’s precipitation. Because
NDVI is also influenced by other factors, the observed values tend to be
scattered around the line with values both above and below the re-
gression line (Fig. 3a). If riparian restoration were unsuccessful (i.e., the
relationship between precipitation and NDVI remains unchanged), the
post-restoration data would continue to fit that pre-restoration pattern
(Fig. 3b). With a successful restoration, the post-restoration NDVI va-
lues would be greater than the values predicted by the pre-restoration
regression (Fig. 3c).
The statistical tests described above provide a quantitative basis for
the assessment of restoration effectiveness. We assessed riparian
M.B. Hausner et al. Ecological Engineering 120 (2018) 432–440

Fig. 3. Synthetic data illustrating the restoration assessment. a) pre-restoration data indicating the linear relationship between annual precipitation and NDVI
(R2 = 0.4). b) sample data showing an ineffective riparian restoration. A t-test shows that these data are not significantly different from the pre-restoration data
(p = 0.74; t-statistic = 0.35). c) sample data showing an effective riparian restoration. A t-test shows that these data are significantly different from the pre-
restoration data (p < 0.005; t-statistic = 3.75).

restoration as effective if two criteria were met: a) the post-restoration vegetation. The strengths of those pre-restoration relationships varied,
NDVI values were greater than the values predicted by the pre-re- likely influenced by topography, soils, local groundwater dynamics and
storation relationship (indicating an increase in photosynthetic poten- other geomorphic features of the study areas. We interpret the slope of
tial, or productivity), and b) the t-test showed a statistically significant the pre-restoration regression equation as representing the strength of
difference between the pre-restoration and post-restoration data (in- precipitation dependence, and the pre-restoration intercept as in-
dicating a change in the functional relationship between precipitation dicative of the availability of alternative sources of water. Lower Camp
and vegetation). Creek, for example, lies alongside the Silvies River and has the greatest
regression intercept (extrapolated NDVI at zero precipitation) of the
3. Results four study areas. The local water table in this study area is higher due to
the proximity of the perennial stream, and the vegetation in this area
For each study area, we found a statistically significant (p < 0.05) can access more water than the vegetation in other study areas. In
linear relationship between pre-restoration precipitation and pre-re- contrast, Flat Creek is the study area most dependent on precipitation,
storation NDVI, although the slopes and intercepts of the various re- with both the greatest sensitivity to precipitation (as indicated by the
lationships varied by a factor of 2 (Table 1; Fig. 4). In each study area, greatest pre-restoration slope) and the least access to other sources of
we also found that post-restoration residuals were statistically distinct water (as indicated by the lowest pre-restoration intercept).
from pre-restoration data (Table 2). Precipitation explains more of the NDVI variance at Flat Creek than any
Fig. 5 shows data from the control areas at Flat Creek and Lower of the other study areas (R2 = 0.521), and the significance of the cor-
Camp Creek, in which 19 years of pre-restoration data were used to relation between precipitation and NDVI is also greatest at Flat Creek
establish the pre-restoration relationship and the following 5 years (still (p < 10−4).
pre-restoration) were tested for a change in that relationship. In this In all four study areas, the post-restoration data were statistically
case, the 19 years of early pre-restoration data (1984–2002) were suf- distinct from the pre-restoration precipitation-NDVI relationship, re-
ficient to establish statistically significant relationships (Table 3a). T- gardless of the pre-restoration correlation, slope, or significance. In all
tests performed on the residuals of five years of late pre-restoration data cases, the post-restoration precipitation data fall within the range of
(2003–2007 control) showed that they were not from a statistically data used to establish the pre-restoration relationship (Fig. 6). Just as
distinct population (Table 3b). In contrast, we found that post-restora- importantly, the control analyses of Flat Creek and Lower Camp Creek
tion data (2012–2017) came from a statistically distinct population did not indicate changes in the riparian community when no restoration
(Table 3c). had occurred. We examined temperature (water year mean tempera-
ture, also acquired from ClimateEngine.org) and autocorrelation as
4. Discussion potential confounding processes. Temperature and NDVI were not sig-
nificantly correlated at any of the study areas – R2 values for the tem-
All four study areas showed statistically significant pre-restoration perature-NDVI relationship ranged from 0.001 to 0.12. Further, there
relationships between precipitation and NDVI, indicating water-limited was no statistically significant autocorrelation (assuming a one-year
lag) within NDVI or precipitation in either the pre-restoration
Table 1 (1984–2007) or post-restoration (2012–2017) periods.
Pre-restoration statistical data (all statistical tests are based on 22 degrees of Instead of simply comparing pre- and post-restoration Landsat
freedom): images to assess a change in vegetation vigor, this method focuses on
Study Area Pre-Restoration Precipitation-NDVI Linear Regression the relationship between NDVI and precipitation and uses all available
Landsat imagery for the time period of interest. Increases in NDVI are
Regression Slope Regression R2 p-value usually associated with increased consumptive water use by riparian
(mm−1) Intercept
vegetation (Groeneveld et al., 2007; Beamer et al., 2013; Nguyen et al.,
Upper Camp Creek 4.00 × 10−4 0.334 0.431 4.93 × 10−4 2015). The statistical analysis presented here shows that this additional
Lower Camp Creek 5.77 × 10−4 0.451 0.288 6.84 × 10−3 water is not provided by precipitation, so it must come from some other
Flat Creek 7.23 × 10−4 0.231 0.521 6.88 × 10−5 source. In the case of Silvies Ranch, the ABDs were installed to raise the
Jack Andy Creek 3.01 × 10−4 0.386 0.337 2.93 × 10−3
level of water in the channel, thereby raising the elevation of the water
table adjacent to the stream channel. The higher water tables result in a

436
M.B. Hausner et al. Ecological Engineering 120 (2018) 432–440

Fig. 4. Pre-restoration (open circles) and post-restoration (filled circles) data for each watershed. In each panel, the best-fit regression line is indicated by the bold
line, and the 95% prediction interval indicated by the solid lines. Pre- and post-restoration statistics for each watershed can be found in Tables 1 and 2, respectively.

Table 2 interrogate the local record from 1984 to the present, establishing a
Post-restoration statistical analysis (all statistical tests based on 5 degrees of baseline condition that is not just a snapshot in time, but a long-term,
freedom): quantifiable relationship that captures a wider range of conditions than
Study Area Post-Restoration Tests of Precipitation-NDVI Regression short-term pre-project monitoring might. This analysis does require
several years of post-restoration data, so it cannot be used as an as-
Statistically significant change? P-value t-statistic sessment tool immediately following the completion of a contemporary
−5 project. However, the method can be applied quickly and inexpensively
Upper Camp Creek Yes 1.32 × 10 16.91
Lower Camp Creek Yes 5.90 × 10−4 7.70 to projects that were implemented as recently as five years ago. As time
Flat Creek Yes 1.46 × 10−5 16.57 passes, the method can also be applied repeatedly to monitor the pro-
Jack Andy Creek Yes 1.88 × 10−3 5.97 ject over the longer timescales (i.e., the decadal monitoring re-
commended by Kondolf and Micheli, 1995) that are seldom considered
in the initial budgeting. Finally, the method could potentially be ap-
plied to consider riparian areas undergoing terrestrialization, although
grouping the pre- and post- terrestrialization data will be more chal-
vadose zone that drains less effectively, increasing the time over which
lenging.
two sources of water (increased soil moisture and higher groundwater)
This statistics-based remote sensing approach is a powerful tool to
are available to the riparian community. The analysis presented here
quantitatively and defensibly assess the effectiveness of one aspect of
offers quantitative and statistically supported evidence for the claim
stream restoration projects. The method does not depend on on-site
that the restoration work effectively achieved that goal.
monitoring and instead uses existing satellite data accessed through
Taking advantage of recent advances in cloud-computing, this
ClimateEngine.org. Climate Engine is designed to make these data-in-
method evaluates riparian restoration at spatial and temporal scales
tensive calculations and extractions accessible to a wide audience of
that have proven challenging in the past (Kondolf and Micheli, 1995;
users, and the analysis presented here requires no specialized skills in
Hobbs and Norton, 1996; Aguiar et al., 2011). Norman et al. (2014)
either remote sensing or Geographic Information System (GIS) pro-
regressed Landsat-derived NDVI against springtime precipitation to
gramming. Climate Engine is currently available at no charge for non-
examine the effects of gabions in cienegas, but limitations in data
commercial activity, making it an attractive option to (usually resource-
processing required them to use a single Landsat image for each year,
limited) NGOs and government agencies that typically sponsor re-
and their nearest usable precipitation gauge was 35 km away from the
storation projects. Similarly, the statistical analysis, although per-
study site. Climate Engine’s easy access to distributed processing of
formed in MATLAB for this paper (see supplemental material for
Landsat images and gridded meteorological data make it possible to

437
M.B. Hausner et al. Ecological Engineering 120 (2018) 432–440

Fig. 5. Control tests on (a) Flat Creek and (b) Lower Camp Creek study areas. The control data (filled black circles) came from the pre-restoration population, but
were not used to establish the pre-restoration regression between precipitation and NDVI.

Table 3
Statistical analyses on the control study areas. (a) pre-restoration (1984–2002)
relationships, (b) control (2003–2007) t-tests, and (c) post-restoration
(2012–2017) t-tests.
(a) Pre-Restoration (1984–2002) Precipitation-NDVI Linear Regression (17 d.f.)
Study Area Regression Slope Regression R2 P-value
Intercept

Flat Creek 6.75 × 10−4 0.242 0.566 2.01 × 10−4

Lower Camp Creek 5.37 × 10−4 0.468 0.311 0.0130

(b) Pre-Restoration Control (2003–2007) Residual Tests for Changes in the

Precipitation-NDVI Regression (4 d.f.)
Study Area Statistically P-value t-statistic
Significant?

Flat Creek no 0.411 0.916

Lower Camp Creek no 0.737 −0.360

(c) Post-Restoration (2012–2017) Residual Tests for Changes in the Precipitation-NDVI

Regression (5 d.f.)
Study Area Statistically P-value t-statistic
Significant? Fig. 6. Pre- and post-restoration precipitation at study areas. The heavy black
−6
line in each column indicates mean pre-restoration precipitation, with the
Flat Creek yes 9.04 × 10 18.27
shaded box indicating ± one standard deviation and the lines indicating the
Lower Camp Creek yes 7.68 × 10−4 7.27
full range (minimum to maximum) of pre-restoration observations. Post-re-
storation precipitation observations are indicated by the open circles. For Lower
Camp Creek and Flat Creek, control data are indicated with filled circles.

program scripts), can be carried out in Excel or in R at limited or no

cost, respectively.
This paper offers a proof-of-concept for a quantitative, statistically
sound assessment of stream and riparian restoration work, but it is also
important to recognize its limitations. It examines only one aspect of
those projects, and it relies on spatially integrated data to do so. The
project must be large enough that data aggregated on a 30 m pixel size
are representative of both the pre-restoration and restored area, and the
spatial scale of the precipitation data must be fine enough that those
data accurately represent the rainfall at the restoration site. The method
identifies changes in the relationship between spatially and temporally
integrated vegetation vigor and precipitation, but it does not identify
the mechanisms by which those changes occurred. Since the method
relies on spatially integrated data, the boundaries of the polygons en-
closing the study areas are particularly important. In this case, the
proof-of-concept goal of this manuscript was achieved by drawing
polygons based on post-restoration riparian areas. In using this method
to examine other projects, however, we recommend drawing the
polygons based on the project goals established before any restoration
work is undertaken. The proof-of-concept here was also demonstrated
in a semi-arid water-limited setting, where there is a marked distinction
in greenness between upland and riparian vegetation. Further testing is
warranted before applying the method in more mesic settings. Finally,
the remote sensing analysis cannot take the place of on-site data col-
lection to evaluate other aspects of the projects, nor can it evaluate the
success of effective riparian restoration in achieving specific project
goals. Wildlife surveys, for example, are still needed to assess the
creation of habitat, and on-site surveys are still required to evaluate
projects intended to halt ongoing incision. However, this method does
offer a way to perform some post-restoration analysis in a data-limited
environment for projects in which pre- and post-restoration on-site data
collection was limited or absent.
5. Conclusions
Using remote sensing and climate data freely available via a cloud-
computing application (ClimateEngine.org) along with relatively
simple statistical analyses, we demonstrate that restoration work at
Silvies Valley Ranch was effective in providing additional water to the
riparian vegetation community, thereby increasing vegetation vigor.

438
M.B. Hausner et al.
We infer that this additional water was provided by shallow ground-
water that was raised by the installation of ABDs along four streams.
The change shown in this study is not merely an observed change in the
vegetation, but a change in the functional relationship between pre-
cipitation and vegetation. The change to that relationship is quantified
and shown to be statistically significant. Although this analysis is lim-
ited to vegetation, it can be performed with freely available data and
common inexpensive software, and it does not depend upon on-site
monitoring, either before or after restoration. The approach and ana-
lysis presented in this paper offers a powerful and cost-effective way to
evaluate effectiveness of one aspect of restoration projects, especially
projects that are implemented with little or no monitoring.
Acknowledgments
The authors thank Scott Campbell for access to Silvies Valley Ranch,
as well as Bruce Wylie (USGS) for his review of the initial manuscript.
This research was supported by the U.S. Department of the Interior
Northwest Climate Science Center; the U.S. Department of Agriculture,
Forest Service, Northwest Regional Climate Hub; the Google Earth
Engine faculty research award program, and U.S. Geological Survey
2012–2017 and 2018–2023 Landsat Science Team funding. Project
sponsors did not participate in the study design, data collection, ana-
lysis, and interpretation, manuscript writing, or the decision to publish
this paper. Any use of trade, product, or firm names is for descriptive
purposes only and does not imply endorsement by the U.S.
Government.
Conflicts of interest
None.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.ecoleng.2018.06.024.
These data include Google maps of the most important areas described
in this article.
References
Abatzoglou, J.T., 2013. Development of gridded surface meteorological data for ecolo-
gical applications and modelling. Int. J. Climatol. 33, 121–131.
Abatzoglou, J.T., Barbero, R., Wolf, J.W., Holden, Z.A., 2014. Tracking interannual
streamflow variability with drought indices in the U.S. Pacific Northwest. J.
Hydrometeorol. 15, 1900–1912.
Aguiar, F.C., Fernandes, M.R., Ferreira, M.T., 2011. Riparian vegetation metrics as tools
for guiding ecological restoration in riverscapes. Knowled. Manage. Aquat.
Ecosyst. 12.
Alexander, G.G., Allan, J.D., 2007. Ecological success in stream restoration: case studies
from the midwestern United States. Environ. Manage. 40, 245–255.
Beamer, J.P., Huntington, J.L., Morton, C.G., Pohll, G.M., 2013. Estimating annual
groundwater evapotranspiration from phreatophytes in the great basin using landsat
and flux tower measurements. J. Am. Water Resour. Assoc. 49, 518–533.
Bernhardt, E.S., Palmer, M.A., Allan, J.D., Alexander, G., Barnas, K., Brooks, S., Carr, J.,
Clayton, S., Dahm, C., Follstad-Shah, J., Galat, D., Gloss, S., Goodwin, P., Hart, D.,
Hassett, B., Jenkinson, R., Katz, S., Kondolf, G.M., Lake, P.S., Lave, R., Meyer, J.L.,
O'Donnell, T.K., Pagano, L., Powell, B., Sudduth, E., 2005. Ecology – Synthesizing US
river restoration efforts. Science 308, 636–637.
Bernhardt, E.S., Sudduth, E.B., Palmer, M.A., Allan, J.D., Meyer, J.L., Alexander, G.,
Follastad-Shah, J., Hassett, B., Jenkinson, R., Lave, R., Rumps, J., Pagano, L., 2007.
Restoring rivers one reach at a time: results from a survey of US river restoration
practitioners. Restor. Ecol. 15, 482–493.
Boudell, J.A., Dixon, M.D., Rood, S.B., Stromberg, J.C., 2015. Restoring functional ri-
parian ecosystems: concepts and applications Preface. Ecohydrology 8, 747–752.
Carroll, R.W., Huntington, J.L., Snyder, K.A., Niswonger, R.G., Morton, C.G., Stringham,
T.K., 2017. Evaluating mountain meadow groundwater response to pinyon-juniper
and temperature in a Great Basin watershed. Ecohydrology.
Daly, C., Halbleib, M., Smith, J.I., Gibson, W.P., Doggett, M.K., Taylor, G.H., Curtis, J.,
Pasteris, P.I., 2008. Physiographically sensitive mapping of climatological tempera-
ture and precipitation across the conterminous United States. Int. J. Climatol.
2008, 34.
Davee, R., Charnley, S., Gosnell, H., 2017. Silvies Valley Ranch, Oregon: Using Artifical
439
Ecological Engineering 120 (2018) 432–440
Beaver Dams to Restore Incised Streams. PNW-RN-577, Pacific Northwest Research
Stations, Portland, OR.
Dawson, S.K., Fisher, A., Lucas, R., Hutchinson, D.K., Berney, P., Keith, D., Catford, J.A.,
Kingsford, R.T., 2016. Remote sensing measures restoration successes, but canopy
heights lag in restoring floodplain vegetation. Rem. Sens. 8, 19.
Dawson, T.E., Pate, J.S., 1996. Seasonal water uptake and movement in root systems of
Australian phraeatophytic plants of dimorphic root morphology: a stable isotope
investigation. Oecologia 107, 13–20.
Devitt, D.A., Fenstermaker, L.F., Young, M.H., Conrad, B., Baghzouz, M., Bird, B.M., 2011.
Evapotranspiration of mixed shrub communities in phreatophytic zones of the Great
Basin region of Nevada (USA). Ecohydrology 4, 807–822.
Essaid, H.I., Hill, B.R., 2014. Watershed-scale modeling of streamflow change in incised
montane meadows. Water Resour. Res. 50, 2657–2678.
Foga, S., Scaramuzza, P.L., Guo, S., Zhu, Z., Dilley, R.D.J., Beckmann, T., Schmidt, G.L.,
Dwyer, J.L., Hughes, M.J., Laue, B., 2017. Cloud detection algorithm comparison and
validation for operational Landsat data products. Remote Sens. Environ. 194,
379–390.
Goetz, S.J., 2006. Remote sensing of riparian buffers: past progress and future prospects.
J. Am. Water Resour. Assoc. 42, 133–143.
Gonzalez del Tanago, M., García de Jalón, D., 2006. Attributes for assessing the en-
vironmental quality of riparian zones. Limnetica 25, 389–402.
Gonzalez, E., Sher, A.A., Tabacchi, E., Masip, A., Poulin, M., 2015. Restoration of riparian
vegetation: a global review of implementation and evaluation approaches in the in-
ternational, peer-reviewed literature. J. Environ. Manage. 158, 85–94.
Goodwin, C.N., Hawkins, C.P., Kershner, J.L., 1997. Riparian restoration in the western
United States: overview and perspective. Restor. Ecol. 5, 4–14.
Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., Moore, R., 2017. Google
Earth Engine: planetary-scale geospatial analysis for everyone. Rem. Sens. Environ.
202, 18–27.
Gregory, S.V., Swanson, F.J., McKee, W.A., Cummins, K.W., 1991. An ecosystem per-
spective of riparian zones. Bioscience 41, 540–551.
Groeneveld, D.P., Baugh, W.M., Sanderson, J.S., Cooper, D.J., 2007. Annual groundwater
evapotranspiration mapped from single satellite scenes. J. Hydrol. 344, 146–156.
Gutman, G., Masek, J.G., 2012. Long-term time series of the Earth's land-surface ob-
servations from space. Int. J. Remote Sens. 33, 4700–4719.
Hansen, M.C., Potapov, P.V., Moore, R., Hancher, M., Turubanova, S.A., Tyukavina, A.,
Thau, D., Stehman, S.V., Goetz, S.J., Loveland, T.R., Kommareddy, A., Egorov, A.,
Chini, L., Justice, C.O., Townshend, J.R.G., 2013. High-resolution Global Maps of
21st-century forest cover change. Science 342, 850–853.
Hobbs, R.J., Norton, D.A., 1996. Towards a conceptual framework for restoration
ecology. Restor. Ecol. 4, 93–110.
Huntington, J., Hegewisch, K., Daudert, B., Morton, C., Abatzoglou, J.T., McEvoy, D.J.,
Erickson, T., 2017. Climate Engine: cloud computing of climate and remote sensing
data for enhanced natural resource monitoring and process understanding. Bull. Am.
Meteorol. Soc. 98, 2397–2410.
Huntington, J., McGwire, K., Morton, C., Snyder, K., Peterson, S., Erickson, T., Niswonger,
R., Carroll, R., Smith, G., Allen, R., 2016. Assessing the role of climate and resource
management on groundwater dependent ecosystem changes in arid environments
with the Landsat archive. Remote Sens. Environ. 185, 186–197.
Huntington, J.L., Niswonger, R.G., 2012. Role of surface-water and groundwater inter-
actions on projected summertime streamflow in snow dominated regions: an in-
tegrated modeling approach. Water Resour. Res. 48, 20.
Kennedy, R.E., Andrefouet, S., Cohen, W.B., Gomez, C., Griffiths, P., Hais, M., Healey,
S.P., Helmer, E.H., Hostert, P., Lyons, M.B., Meigs, G.W., Pflugmacher, D., Phinn,
S.R., Powell, S.L., Scarth, P., Sen, S., Schroeder, T.A., Schneider, A., Sonnenschein, R.,
Vogelmann, J.E., Wulder, M.A., Zhu, Z., 2014. Bringing an ecological view of change
to Landsat-based remote sensing. Front. Ecol. Environ. 12, 339–346.
Kennedy, R.E., Yang, Z., Gorelick, N., Braaten, J., Cavalcante, L., Cohen, W.B., Healey, S.,
2018. Implementation of the LandTrendr algorithm on Google Earth Engine. Rem.
Sens. 10, 691.
Kondolf, G.M., 1995. Elements for effective evaluation of stream restoration. Restor. Ecol.
3, 133–136.
Kondolf, G.M., Anderson, S., Lave, R., Pagano, L., Merenlender, A., Bernhardt, E.S., 2007.
Two decades of river restoration in California: what can we learn? Restor. Ecol. 15,
516–523.
Kondolf, G.M., Micheli, E.R., 1995. Evaluating stream restoration projects. Environ.
Manage. 19, 1–15.
Lake, P.S., Bond, N., Reich, P., 2007. Linking ecological theory with stream restoration.
Freshw. Biol. 52, 597–615.
Lawley, V., Lewis, M., Clarke, K., Ostendorf, B., 2016. Site-based and remote sensing
methods for monitoring indicators of vegetation condition: an Australian review.
Ecol. Ind. 60, 1273–1283.
Loheide, S.P., Gorelick, S.M., 2005. A local-scale, high-resolution evapotranspiration
mapping algorithm (ETMA) with hydroecological applications at riparian meadow
restoration sites. Remote Sens. Environ. 98, 182–200.
McEvoy, D.J., Huntington, J.L., Abatzoglou, J.T., Edwards, L.M., 2012. An evaluation of
multiscalar drought indices in Nevada and Eastern California. Earth Interact 16, 18.
McGwire, K., Minor, T., Fenstermaker, L., 2000. Hyperspectral mixture modeling for
quantifying sparse vegetation cover in arid environments. Remote Sens. Environ. 72,
360–374.
Mitchell, K.E., Lohmann, D., Houser, P.R., Wood, E.F., Schaake, J.C., Robock, A.,
Cosgrove, B.A., Sheffield, J., Duan, Q.Y., Luo, L.F., Higgins, R.W., Pinker, R.T.,
Tarpley, J.D., Lettenmaier, D.P., Marshall, C.H., Entin, J.K., Pan, M., Shi, W., Koren,
V., Meng, J., Ramsay, B.H., Bailey, A.A., 2004. The multi-institution North American
Land Data Assimilation System (NLDAS): utilizing multiple GCIP products and
partners in a continental distributed hydrological modeling system. J. Geophys. Res.-
M.B. Hausner et al.
Atmos. 109, 32.
Naiman, R.J., Decamps, H., 1997. The ecology of interfaces: riparian zones. Annu. Rev.
Ecol. Syst. 28, 621–658.
Naiman, R.J., Decamps, H., Pollock, M., 1993. The role of riparian corridors in main-
taining regional biodiversity. Ecol. Appl. 3, 209–212.
Nguyen, U., Glenn, E.P., Nagler, P.L., Scott, R.L., 2015. Long-term decrease in satellite
vegetation indices in response to environmental variables in an iconic desert riparian
ecosystem: the Upper San Pedro, Arizona, United States. Ecohydrology 8, 610–625.
Norman, L., Villarreal, M., Pulliam, H.R., Minckley, R., Gass, L., Tolle, C., Coe, M., 2014.
Remote sensing analysis of riparian vegetation response to desert marsh restoration
in the Mexican Highlands. Ecol. Eng. 70, 241–254.
NRC, 2002. Riparian Areas: Functions and Strategies for Management. The National
Academies Press, Washington, DC.
Orellana, F., Verma, P., Loheide, S.P., Daly, E., 2012. Monitoring and modeling water-
vegetation interactions in groundwater-dependent ecosystems. Rev. Geophys. 50, 24.
Ozesmi, S.L., Bauer, M.E., 2002. Satellite remote sensing of wetlands. Wetlands Ecol.
Manage. 10, 381–402.
Palmer, M.A., Bernhardt, E.S., Allan, J.D., Lake, P.S., Alexander, G., Brooks, S., Carr, J.,
Clayton, S., Dahm, C.N., Shah, J.F., Galat, D.L., Loss, S.G., Goodwin, P., Hart, D.D.,
Hassett, B., Jenkinson, R., Kondolf, G.M., Lave, R., Meyer, J.L., O'Donnell, T.K.,
Pagano, L., Sudduth, E., 2005. Standards for ecologically successful river restoration.
J. Appl. Ecol. 42, 208–217.
Patten, D.T., 1998. Riparian ecosystems of semi-arid North America: diversity and human
impacts. Wetlands 18, 498–512.
Pekel, J.F., Cottam, A., Gorelick, N., Belward, A.S., 2016. High-resolution mapping of
global surface water and its long-term changes. Nature 540, 418-+.
Perry, L.G., Andersen, D.C., Reynolds, L.V., Nelson, S.M., Shafroth, P.B., 2012.
Vulnerability of riparian ecosystems to elevated CO2 and climate change in arid and
semiarid western North America. Glob. Change Biol. 18, 821–842.
Pilliod, D.S., Rohde, A.T., Charnley, S., Davee, R.R., Dunham, J.B., Gosnell, H., Grant,
G.E., Hausner, M.B., Huntington, J.L., Nash, C., 2018. Survey of Beaver-related re-
storation practices in rangeland streams of the Western USA. Environ. Manage. 61,
58–68.
Rigge, M., Smart, A., Wylie, B., 2013a. Optimal placement of off-stream water sources for
ephemeral stream recovery. Rangeland Ecol. Manage. 66, 479–486.
Rigge, M., Wylie, B., Gu, Y.X., Belnap, J., Phuyal, K., Tieszen, L., 2013b. Monitoring the
status of forests and rangelands in the Western United States using ecosystem per-
formance anomalies. Int. J. Remote Sens. 34, 4049–4068.
Ruiz-Jaen, M.C., Aide, T.M., 2005. Restoration success: how is it being measured? Restor.
Ecol. 13, 569–577.
Sabo, J.L., Sponseller, R., Dixon, M., Gade, K., Harms, T., Heffernan, J., Jani, A., Katz, G.,
Soykan, C., Watts, J., Welter, A., 2005. Riparian zones increase regional species
440
Ecological Engineering 120 (2018) 432–440
richness by harboring different, not more, species. Ecology 86, 56–62.
Smith, A.M.S., Kolden, C.A., Tinkham, W.T., Talhelm, A.F., Marshall, J.D., Hudak, A.T.,
Boschetti, L., Falkowski, M.J., Greenberg, J.A., Anderson, J.W., Kliskey, A., Alessa, L.,
Keefe, R.F., Gosz, J.R., 2014. Remote sensing the vulnerability of vegetation in nat-
ural terrestrial ecosystems. Remote Sens. Environ. 154, 322–337.
Stromberg, J.C., Beauchamp, V.B., Dixon, M.D., Lite, S.J., Paradzick, C., 2007.
Importance of low-flow and high-flow characteristics to restoration of riparian ve-
getation along rivers in and south-western United States. Freshw. Biol. 52, 651–679.
Trowbridge, W.B., 2007. The role of stochasticity and priority effects in floodplain re-
storation. Ecol. Appl. 17, 1312–1324.
USGS, 2018. Landsat 4-7 Surface Reflectance (LEDAPS) Product. U.S. Department of the
Interior, U.S. Geological Survey.
USGS, 2018. Landsat 8 Surface Reflectance Code (LaSRC) Product. U.S. Department of the
Interior, U.S. Geological Survey.
Walker, L.R., Walker, J., del Moral, R., 2007. Forging a new alliance between succession
and restoration. In: Walker, L.R., Walker, J., Hobbs, R.J. (Eds.), Linking Restoration
and Ecological Succession. Springer Science and Business Media, New York, NY, USA,
pp. 1–18.
Weisberg, P.J., Mortenson, S.G., Dilts, T.E., 2013. Gallery forest or herbaceous wetland?
The need for multi-target perspectives in riparian restoration planning. Restor. Ecol.
21, 12–16.
Wulder, M.A., Masek, J.G., Cohen, W.B., Loveland, T.R., Woodcock, C.E., 2012. Opening
the archive: how free data has enabled the science and monitoring promise of
Landsat. Rem. Sens. Environ. 122, 2–10.
Wylie, B., Rigge, M., Brisco, B., Murnaghan, K., Rover, J., Long, J., 2014. Effects of dis-
turbance and climate change on ecosystem performance in the yukon river basin
boreal forest. Rem. Sens. 6, 9145–9169.
Wylie, B.K., Boyte, S.P., Major, D.J., 2012. Ecosystem performance monitoring of ran-
gelands by integrating modeling and remote sensing. Rangeland Ecol. Manage. 65,
241–252.
Wylie, B.K., Zhang, L., Bliss, N., Ji, L., Tieszen, L.L., Jolly, W.M., 2008. Integrating
modelling and remote sensing to identify ecosystem performance anomalies in the
boreal forest, Yukon River Basin, Alaska. Int. J. Digital Earth 1, 196–220.
Xia, Y.L., Mitchell, K., Ek, M., Sheffield, J., Cosgrove, B., Wood, E., Luo, L.F., Alonge, C.,
Wei, H.L., Meng, J., Livneh, B., Lettenmaier, D., Koren, V., Duan, Q.Y., Mo, K., Fan,
Y., Mocko, D., 2012. Continental-scale water and energy flux analysis and validation
for the North American Land Data Assimilation System project phase 2 (NLDAS-2): 1.
Intercomparison and application of model products. J. Geophys. Res.-Atmos.
117 (27).
Zhu, Z., Woodcock, C.E., 2012. Object-based cloud and cloud shadow detection in
Landsat imagery. Remote Sens. Environ. 118, 83–94.